3D P-wave velocity model of the Alaska Peninsula subduction zone from the AACSE active-source data set

This is a 3D, P-wave, crustal-scale velocity model of the Alaska subduction zone that encompasses both the incoming and overriding plates and spans the Semidi segment and the SW Kodiak aspertiy. This model was derived from the traveltime inversion of first-arrivals interpreted from active-source ocean bottom seismometer data and onshore seismometer data acquired as part of the Alaska Amphibious Community Seismic Experiment (AACSE; MGL1903). The model also incorporates data from the 2D OBS profile AL03 acquired as part of the Alaska Langseth Experiment to Understand the MegaThrust (ALEUT; MGL1110) and vertical-component data from TA and AEC stations located on the Alaska Peninsula, Kodiak Island (S19K), and the smaller Sitkinak (SII), Chirikof (CHI), and Shumagin Islands (CHN). We interpreted 42,103 first-arrivals out of 609,915 available traces with 32,024 (76%) coming from 37 of the AACSE instruments, 3,541 (8%) from the 3 AEC instruments, and 6,538 (16%) from the 20 instruments in the AL03 data set. The seismic velocity model file is in ASCII CSV format with a header for each column. The seismic velocity model contains six columns as follows: longitude, latitude, Z in km, model seafloor depth in km, P-wave velocity in km/s, and standard deviation in km/s. The sea surface is set to Z=0, meaning negative Z values indicate topography or air (i.e., above sea level). The model coordinates have been rotated and converted back to latitude and longitude. The model has been regridded to a uniform spacing of 0.25 degrees in lat/lon and 0.5 km in depth. The dimensions of the model are 227 by 446 by 86 in latitude, longitude, and depth (km), respectively. The water velocity is set to 1.5 km/s and the air velocity is 0.33 km/s. Voxels outside of the original rotated model space contain NaN values for P-wave velocity and standard deviation. The starting P-wave velocity model was created from the results of a 2D traveltime inversion of the 2D AL03 data. To create the final model, we performed a Monte Carlo analysis by creating 50 perturbed starting models, adding random Gaussian noise to the traveltime data, and inverting the data with the same forward calculation and regularization parameters. The final Vp model is the average of the 50 ensembles, which has a mean standard deviation of 0.041 km/s within the overlapping regions of ray coverage. The associated navigation files contain the source and receiver locations used in this study. Funding to create this velocity model was provided through NSF grants OCE-1947758 and OCE-1948087.